/NetDissect-Lite

Light version of Network Dissection for Quantifying Interpretability of Networks

Primary LanguageJupyter Notebook

Network Dissection Lite in PyTorch

Introduction

This repository is fork of NetDissect, which contains the demo code for the work Network Dissection: Quantifying Interpretability of Deep Visual Representations. We also test the Network Dissection results on ResNet18 and ResNet18+CBAM. This code is written in pytorch and python3.6, tested on Ubuntu 20.04(google colab).

Originally written by: David Bau∗ , Bolei Zhou∗ , Aditya Khosla, Aude Oliva, and Antonio Torralba

Reimplemented and extended by: Ruifeng Xu, Yuhang Mei

Classification for ResNet18 and ResNet18+CBAM on CIFAR and ImageNet

  • Go to the 'Classification' folder
  • Run in google colab:
    • TrainingOnCifar10Dataset - CIFAR10
    • TrainingOnCifar100Dataset - CIFAR100
    • TrainingOnImageNet - ImageNet

Run NetDissect for ResNet18 and ResNet18+CBAM

  • Go to the 'ND' folder

  • Run in google colab

Reference

@inproceedings{netdissect2017,
  title={Network Dissection: Quantifying Interpretability of Deep Visual Representations},
  author={Bau, David and Zhou, Bolei and Khosla, Aditya and Oliva, Aude and Torralba, Antonio},
  booktitle={Computer Vision and Pattern Recognition},
  year={2017}
}